Consumer Perspectives on. Insurance Credit Scoring and Disparate Impact Standard for Unfair Discrimination. March 2013

Transcription

1 Consumer Perspectives on Insurance Credit Scoring and Disparate Impact Standard for Unfair Discrimination Casualty Actuarial Society Ratemaking and Product Management Seminar March 2013 Birny Birnbaum Center for Economic Justice

2 The Center for Economic Justice CEJ is a non-profit consumer advocacy organization dedicated to representing the interests of low-income and minority consumers as a class on economic justice issues. Most of our work is before administrative agencies on insurance, financial services and utility issues. On the Web: Slide 2 Consumer Perspectives Insurance Scoring, Disparate Impact

3 Why CEJ Works on Insurance Issues Essential Financial Security Tool for Individual and Community Economic Development: CEJ Works to Ensure Access and Fair Prices for These Essential Products and Services, particularly for Low- and Moderate-Income Consumers. Primary Institution to Promote Loss Prevention and Mitigation: CEJ Works to Ensure Insurance Institutions Maximize Their Role in Efforts to Reduce Loss of Life and Property from Catastrophic Events. Slide 3 Consumer Perspectives Insurance Scoring, Disparate Impact

4 Overview 1. Insurance Credit Scoring (CS) Is Inherently Unfair 2. CS Has A Disparate Impact on Low-Income and Minority Consumers; CS Reflects and Perpetuates Historical Inequities. 3. Consumer Credit Data Quality, Data Variability and Difficulty for Consumers to Fix Errors Are Additional Concerns. 4. CS Undermines Core Public Policy Goals of Insurance: Universal Coverage and Loss Mitigation. 5. Department of Housing and Urban Development s Recent Rule Regarding Disparate Impact Covers Residential Property Insurance. Slide 4 Consumer Perspectives Insurance Scoring, Disparate Impact

5 Insurance Credit Scoring is Inherently Unfair Penalizes Victims of Medical, Economic Catastrophes Penalizes Consumers for Abusive Lending Practices / Broader Economic Conditions Arbitrary and Illogical Results Unrelated to How Well a Consumer Manages Her Finances Slide 5 Consumer Perspectives Insurance Scoring, Disparate Impact

6 Consumers Hammered By Financial Crisis and Recession Reckless and Abusive Lending High Unemployment Wage Cuts Credit Limit Reductions Increases in Loan and Credit Card Fees Increasing Medical Costs Record or Near-Record Highs in Delinquencies Foreclosures Bankruptcies Following Charts from Slide 6 Consumer Perspectives Insurance Scoring, Disparate Impact

7 Slide 7 Consumer Perspectives Insurance Scoring, Disparate Impact

8 Slide 8 Consumer Perspectives Insurance Scoring, Disparate Impact

9 Slide 9 Consumer Perspectives Insurance Scoring, Disparate Impact

10 Slide 10 Consumer Perspectives Insurance Scoring, Disparate Impact

11 Slide 11 Consumer Perspectives Insurance Scoring, Disparate Impact

12 Slide 12 Consumer Perspectives Insurance Scoring, Disparate Impact

13 BI Pure Premium: Nevada & Multi-State Slide 13 Consumer Perspectives Insurance Scoring, Disparate Impact

14 PD Pure Premium: Nevada & Multi-State l Slide 14 Consumer Perspectives Insurance Scoring, Disparate Impact

15 Collision Pure Premium: Nevada & Multi-State Slide 15 Consumer Perspectives Insurance Scoring, Disparate Impact

16 BI Pure Premium: Florida & Multi-State Slide 16 Consumer Perspectives Insurance Scoring, Disparate Impact

17 PD Pure Premium: Florida & Multi-State Slide 17 Consumer Perspectives Insurance Scoring, Disparate Impact

18 Collision Pure Premium: Florida & Multi-State Slide 18 Consumer Perspectives Insurance Scoring, Disparate Impact

19 Causes of Bankruptcies Harvard Study of Bankruptcies in 2001: 87% of Bankruptcies Caused by Job Loss, Medical Bills or Divorce 46.2% from Medical Problems Harvard Study of Bankruptcies in 2007: 62.1% of Bankruptcies Caused by Medical Problems 75% of These Were Families With Health Insurance. Slide 19 Consumer Perspectives Insurance Scoring, Disparate Impact

20 Consumer Credit Data Quality Issues February 2012 Federal Trade Commission Study the study design called for consumers to be randomly selected from the population of interest (consumers with credit histories at the three national CRAs). Ultimately, 1,001 study participants reviewed 2,968 credit reports (roughly three per participant) with a study associate who helped them identify potential errors. Slide 20 Consumer Perspectives Insurance Scoring, Disparate Impact

21 2013 Federal Trade Commission Study: Findings One in four consumers identified errors on their credit reports that might affect their credit scores; One in five consumers had an error that was corrected by a credit reporting agency (CRA) after it was disputed, on at least one of their three credit reports; Four out of five consumers who filed disputes experienced some modification to their credit report; Slightly more than one in 10 consumers saw a change in their credit score after the CRAs modified errors on their credit report; and Approximately one in 20 consumers had a maximum score change of more than 25 points and only one in 250 consumers had a maximum score change of more than 100 points. Slide 21 Consumer Perspectives Insurance Scoring, Disparate Impact

22 2013 Federal Trade Commission Study Findings Understate Number of Errors and Difficulty for Consumers to Fix Errors Credit Scores impacted as much or more by absence of information than by presence of false information. Consumers are much less likely to identify absence of information as an error. Automated Credit Report Dispute System: Credit bureau often limits its role in disputes to little more than assigning codes as to what type of dispute is at issue: the credit bureaus do not examine documents, contact consumers by phone or , or exercise any form of human discretion in resolving a dispute. The vast majority (85%) of credit reporting disputes are passed on to the company (known as a furnisher) that provided the information. Slide 22 Consumer Perspectives Insurance Scoring, Disparate Impact

23 Consumer Financial Protection Bureau Report, Dec 2012 More than half of the trade lines in the credit bureau databases are supplied by the credit card industry: Credit reporting companies get their information from a variety of industries but more than half of the account information is supplied by credit card companies. Specifically, 40 percent comes from bank cards, such as general credit cards, and 18 percent comes from retail credit cards. Only 7 percent comes from mortgage lenders or servicers, and only 4 percent comes from auto lenders. Slide 23 Consumer Perspectives Insurance Scoring, Disparate Impact

24 Consumer Financial Protection Bureau Report More than a third of disputes have to do with collections: In 2011, consumers reached out to the credit reporting companies roughly 8 million times, resulting in disputes of 32 to 38 million items in their credit files. Almost 40 percent of the disputes relate to debt in collections, and debt in collections is five times more likely to be disputed than mortgage information. According to the industry, some of this may have to do with consumers incentive to dispute any negative information on their reports. Slide 24 Consumer Perspectives Insurance Scoring, Disparate Impact

25 Consumer Financial Protection Bureau Report Fewer than one in five people obtain copies of their credit report each year: The most effective way for consumers to identify errors in their reports is to obtain copies and review them. But only about 44 million consumers per year, or about one in five, obtain copies of their files. Most information contained in credit reports comes from a few large companies: Most information contained in credit files comes from a small number of large banks and other financial institutions. In fact, the top 10 data furnishers provide 57 percent of the trade lines coming into the credit reporting companies. The top 50 furnishers provide 72 percent. And the top 100 furnishers provide 76 percent. Slide 25 Consumer Perspectives Insurance Scoring, Disparate Impact

26 Consumer Financial Protection Bureau Report Most complaints are forwarded to the furnishers that provided the original information: The credit reporting companies resolve an average of 15 percent of consumer disputed items internally, without getting the data furnishers involved. The remaining 85 percent are passed on to the furnishers. Today s report, however, found that the documentation consumers mail in to support their cases may not be getting passed on to the data furnishers for them to properly investigate and report back to the credit reporting company. Slide 26 Consumer Perspectives Insurance Scoring, Disparate Impact

27 Consumer Credit Data Quality Issues Scores Vary Significantly by Geographic Area Differences correlated with income levels, unemployment rates San Jose-Sunnyvale-Santa Clara, CA. 700 San Francisco-Oakland-Fremont, CA 696 Madison, WI. 694 Honolulu 693 Minneapolis-St. Paul-Bloomington, MN- WI. 691 Bridgeport-Stamford-Norwalk, CT. 690 Boston-Cambridge-Quincy, MA-NH 689 Oxnard-Thousand Oaks-Ventura, CA 685 Portland-South Portland-Biddeford, ME 685 Seattle-Tacoma-Bellevue, WA 685 TransUnion, February 7, 2013 Memphis, TN-MS-AR. 638 McAllen-Edinburg-Mission, TX 639 Jackson, MS. 642 El Paso, TX 650 Columbia, SC 650 Las Vegas-Paradise, NC 650 Little Rock-North Little Rock-Conway, AR 651 Baton Rouge, LA 651 Lakeland-Winter Haven, FL. 651 Augusta-Richmond County, GA-SC 651 Slide 27 Consumer Perspectives Insurance Scoring, Disparate Impact

28 Scores Change Significantly Over Short Periods of Time Roughly 70 percent of credit scores change by up to 20 points in any given 90-day window. Most consumers experience a score improvement rather than a score drop, with 56 percent of consumers shifting higher, 34 percent shifting lower and 10 percent staying the same. Experian Credit Cornerstone, January 23, 2013 Slide 28 Consumer Perspectives Insurance Scoring, Disparate Impact

29 Insurance Credit Scoring Not Objective Differences across credit bureaus Differences within a credit bureau due to lender choices Changes in definitions of credit report items bankruptcy law change Public policy initiatives changing credit scores moratorium on foreclosures Timing of report balance to limits varies by time of the month Decisions of lenders not reporting limits, changing limits Slide 29 Consumer Perspectives Insurance Scoring, Disparate Impact

30 Insurance Credit Scoring Is Subject to Manipulation Invitations/Solicitations for Manipulation Piggy-Back on another consumer Shift balances from one car to multiple cards Penalizes Consumer for Rational Behavior Shop around for best rates Cancel a card when lender acts unfairly Get a card to get 10% first visit discount Slide 30 Consumer Perspectives Insurance Scoring, Disparate Impact

31 Correlation to Race and Income The Missouri DOI Study The insurance credit-scoring system produces significantly worse scores for residents of high-minority ZIP Codes. The insurance credit-scoring systems produce significantly worse scores for residents of low-income ZIP Code. The relationship between minority concentration in a ZIP Code and credit scores remained after eliminating a broad array of socioeconomic variables, such as income, educational attainment, marital status and unemployment rates, as possible causes. Indeed, minority concentration proved to be the single most reliable predictor of credit scores. Slide 31 Consumer Perspectives Insurance Scoring, Disparate Impact

32 Correlation to Race and Income The Texas DOI Study The individual policyholder data shows a consistent pattern of differences in credit scores among the different racial/ethnic groups. The average credit scores for Whites and Asians are better than those for Blacks and Hispanics. In addition, Blacks and Hispanics tend to be over-represented in the worse credit score categories and under-represented in the better credit score categories. Slide 32 Consumer Perspectives Insurance Scoring, Disparate Impact

33 The Flawed Federal Trade Commission Study Failed to obtain a comprehensive and independent data set for analysis. One FTC Commissioner voted against issuing the study out of concern for quality of study. FTC has since changed approach for homeowners analysis. Even with limited data, FTC found racial impact. Data on scores of applicants as opposed to data only for policyholders would show much greater disparate impact. Slide 33 Consumer Perspectives Insurance Scoring, Disparate Impact

34 Credit Scoring Reflects / Perpetuates Historical Inequities Segregation therefore racialized and intensified the consequences of the American housing bubble. Hispanic and black home owners, not to mention entire Hispanic and black neighborhoods, bore the brunt of the foreclosure crisis. This outcome was not simply a result of neutral market forces but was structured on the basis of race and ethnicity through the social fact of residential segregation. Rugh and Massey, Racial Segregation and the American Foreclosure Crisis, American Sociological Review, Vol 75, No 5 Slide 34 Consumer Perspectives Insurance Scoring, Disparate Impact

35 Undermines the Core Public Policy Goals of Insurance Undermines the goal of universal coverage by worsening the availability and affordability of insurance for those consumers with the least means to purchase insurance; and Undermines the loss mitigation role of insurance by o Placing great emphasis on a rating factor which has no ability to promote loss mitigation by policyholder; and o Encouraging consumers to spend time manipulating credit scores instead of true loss mitigation activities. Slide 35 Consumer Perspectives Insurance Scoring, Disparate Impact

36 Insurance Scoring Is Not Needed States Which Ban Insurance Credit Scoring, including California and Massachusetts Have Thriving Markets. Insurers Entered The Massachusetts Auto Market After Partial Deregulation, Even Though Insurance Credit Scoring Is Banned. Insurance Credit Scoring Not Needed to Avoid Adverse Selection. Insurance Credit Scoring Not Needed With Modern Risk Classification. Slide 36 Consumer Perspectives Insurance Scoring, Disparate Impact

37 Claims of Consumer Benefits of Insurance Scoring Are Refuted by Objective, Independent Data Allows Insurers to Write More Business Fact: Uninsured Motorist Rate Has Increased Countrywide While Uninsured Motorist Rate Has Declined in CA and MA where Insurance Credit Scoring is Banned Fact: Auto Residual Market Has Declined More in CA than Countrywide Fact: Creditor-Placed (Force-Placed) Insurance Has Skyrocketed in Past 5 Years No Objective Evidence to Support This Claim Slide 37 Consumer Perspectives Insurance Scoring, Disparate Impact

38 Industry Claim: Insurance Credit Scores Reflect Personal Responsibility Blaming the Victim Claim is Factually Incorrect Actual Causes of Financial Distress Typically Beyond Control of Consumers Traditional Credit Reports Missing Information on Financial Responsibility, Let Alone Personal Responsibility Recent Actions by Credit Scoring Modelers to Utilize Non- Traditional Credit Information Documents Disparate Impact of Traditional Credit Information on Low-Income and Minority Consumers. Slide 38 Consumer Perspectives Insurance Scoring, Disparate Impact

39 Unfair Discrimination: Intentional vs. Disparate Impact Actuarial Principles and Standards of Practice primary limitations on risk classifications are statistical relationship to expected costs and statutory or regulatory prohibitions. If race were not a prohibited factor and insurers found a statistical relationship between race and likelihood of claims, insurers would be compelled to use race. Public policy prohibits use of race regardless of any statistical relationship to claims. Some states prohibit use of other characteristics, including consumer credit information, gender and being a legislator. Slide 39 Consumer Perspectives Insurance Scoring, Disparate Impact

40 Department of Housing and Urban Development Disparate Impact Rule (a) Discriminatory effect. A practice has a discriminatory effect where it actually or predictably results in a disparate impact on a group of persons or creates, increases, reinforces, or perpetuates segregated housing patterns because of race, color, religion, sex, handicap, familial status, or national origin. (b) Legally sufficient justification. (1) A legally sufficient justification exists where the challenged practice: (i) Is necessary to achieve one or more substantial, legitimate, nondiscriminatory interests of the respondent... and (ii) Those interests could not be served by another practice that has a less discriminatory effect. Slide 40 Consumer Perspectives Insurance Scoring, Disparate Impact

41 HUD Disparate Impact Rule: Insurance Trades Apoplectic NAMIC says that, if allowed to stand, this rule could undermine the entire process of insurance underwriting, effectively blinding insurers as they attempt to determine potential risks and appropriate pricing, and needlessly raising costs for all consumers." [NAMIC] interprets the rule as codifying the use of disparateimpact analysis to prove allegations of unlawful discrimination with regards to homeowners insurance, meaning any factor used by insurers to assess risk could be challenged if it produces statistically disproportionate outcomes among particular demographic groups. Slide 41 Consumer Perspectives Insurance Scoring, Disparate Impact

42 HUD Disparate Impact Rule: Insurance Trades Apoplectic Leon Buck, PCI s assistant vice president, federal government relations, notes, States already prohibit discriminatory practices and have comprehensive enforcement, but the HUD rule puts in jeopardy the use of longstanding, sound, state-approved actuarial factors that are the foundation of responsible insurance underwriting. NAMIC: Businesses should not be penalized "because of a statistical disparity," "As long as it could be shown that there was no intent to discriminate racially or ethnically, there should be no controversy." Slide 42 Consumer Perspectives Insurance Scoring, Disparate Impact

43 Example of Disparate Impact Complaint: National Fair Housing Alliance against [Company X] [Company X] refuses to insure homes with flat roofs in and around Wilmington, DE. NFHA sent testers seeking insurance quotes for homes in the Wilmington area to six agencies, two of which were [Company X]. [Company X] agents refused, indicating [Company X] does not insure homes with flat roofs. Other non-[company X] agencies offered quotes. NFHA referenced research by University of Delaware Center for Community Research and Service. Research shows very high correlation of percentage of flat roofs in a given area and the percentage of African American or minority owner-occupied units in the area. Slide 43 Consumer Perspectives Insurance Scoring, Disparate Impact

44 NFHA Disparate Impact Complaint 2 CCRS s study employed geographic information systems and statistical software to investigate the correlation between race and roof type at the census tract level (by considering the percentage of minority households and the percentage of homes with flat roofs within a given census tract). Demographic and other housing data were obtained from the 2000 Census of Population and Housing. Data on roof type was based on tax parcel information, which included both information on roof type and grid number (a defined, geographic area that is larger than a block but smaller than a census tract). There were a total of 648 grids, with 164,963 dwelling units, that had roof type data within the relevant area. Slide 44 Consumer Perspectives Insurance Scoring, Disparate Impact

45 NFHA Disparate Impact Complaint 3 Correlation of Key Racial Variables with Percentage of Flat Roof Owner-Occupied Dwelling Units by Census Tract Pearson Variable Correlation Significance Dwelling units owned and occupied by African-American and non- Hispanic households as a percent of all owner-occupied dwelling unit Dwelling units owned and occupied by White and non-hispanic households as a percent of all owneroccupied dwelling units Slide 45 Consumer Perspectives Insurance Scoring, Disparate Impact

46 NFHA Disparate Impact Complaint 4 Simple Bivariate Regression of Percentage of Owner-Occupied Flat Rood Dwellings and Race/Ethnicity of Housing Unit Households by Census Tract Adjusted R2 Variable Coefficient Dwelling units owned and occupied by African-American and non-hispanic households as a percent of all owneroccupied dwelling unit Dwelling units owned and occupied by White and non-hispanic households as a percent of all owner-occupied dwelling units Slide 46 Consumer Perspectives Insurance Scoring, Disparate Impact

47 NFHA Disparate Impact Complaint 5 Additional analyses demonstrated that homes within Wilmington proper have a disproportionate share of flat roof residences as compared to non-wilmington census tracts. The researchers inserted a variable to indicate whether a census tract was inside or outside of Wilmington. The adjusted R² value for each was over 0.84, which establishes that location inside or outside of Wilmington and the race/ethnicity of owner-occupied households together accounted for over 84% of the variation of the percentage of owner-occupied flat roof dwelling units in a census tract. Wilmington s population was 54% African-American and 36% white as of the 2000 Census (the data relied upon in the study). Slide 47 Consumer Perspectives Insurance Scoring, Disparate Impact

48 NFHA Disparate Impact Complaint 6 The statistical evidence proves that [Company X] s policy disproportionately impacts African-American and minority communities, as there is a statistically significant relationship between minority populations and homes that have flat roofs. Coupled with NFHA s testing, which illustrates that [Company X] refuses to insure homes with flat roofs in Wilmington, Delaware and surrounding areas, Complainant has established a prima facie case that [Company X] s no flat roof policy has a racially disparate impact on African-American and minority communities in violation of the Fair Housing Act. Slide 48 Consumer Perspectives Insurance Scoring, Disparate Impact

49 NFHA Disparate Impact Complaint 7 NFHA s testing also reveals the absence of any business justification for [Company X] s practice of refusing to write insurance policies for homes with flat roofs in Wilmington and surrounding areas. [Company X] agents did not disclose any business reason for [Company X] s refusal to write homes with flat roofs. Other insurers offered coverage for homes with flat roofs. Finally, NFHA tested [Company X] insurance companies in Washington, DC and Philadelphia, Pennsylvania. NFHA testers, purporting to seek information about homeowners insurance for homes with flat roofs, each contacted two, separate [Company X] insurance companies in each jurisdiction. The investigation revealed that [Company X] does not have a policy against insuring homes with flat roofs in these cities. Slide 49 Consumer Perspectives Insurance Scoring, Disparate Impact

50 Disparate Impact: A Statistical Test Appropriate for Insurance If a statistical relationship is justification for use of a risk classification, it makes sense that the same statistical relationship test is a valid measure of discrimination on the basis of protected classes. As a public policy matter, if insurers are prohibited from intentionally discriminating on the basis of race, it follows that non-intentional discrimination that has the effect of discriminating on the basis of race is also prohibited unless there is a business justification and no less discriminatory alternative. How Can Actuaries Identify and Mitigate Disparate Impact? Slide 50 Consumer Perspectives Insurance Scoring, Disparate Impact